JPS61246843A - Multiple job start control system - Google Patents

Abstract

PURPOSE:To reduce the start processing time by setting the environment close to the optimum state (a state that the entire processing time is equal to that of CPU processing) by sub-schedule information to start automatically the user job. CONSTITUTION:The job schedule information 1 is registered on a system as a part of so-called procedure and plural inputted jobs are registered on a magnetic disc device (DASD). A scheduler 3 reads the job schedule information 1 and an analysis section 4 analyzes the control sentence and decides the start control timing with respect to the designated job, sequentially start job name possible for multiplex processing, start waiting designation of succeeding job for a prescribed time, designation of job start end condition just before the succeeding job and end of schedule.

Description

[Detailed Description of the Invention] [Summary] When starting a plurality of jobs in succession, the degree of job multiplicity is always controlled to be in an efficient and optimal state that is neither excessive nor insufficient.

[Industrial application field]

The present invention relates to a job startup control method in a data processing system, and particularly to a control method for starting a plurality of jobs at once before starting a job.

[Conventional technology]

For example, in a data processing system, a group of jobs for performing a certain task are started all at once when the system is started, and the task is started when all of those jobs have finished starting. ing. FIG. 3 exemplarily shows this.

Generally speaking, there are two methods for starting multiple jobs in succession: a method for starting multiple jobs individually in series, and a method for starting multiple jobs in parallel.

In the case of serial startup, the next job is started after the startup of one job is completed, so the total startup processing time is the startup processing time of each job (the startup processing time of a single job). ) will require the total time.

In addition, in the case of a single job startup process, the CPU wait time increases when executing the I10 instruction, etc., and the processing efficiency of the CPU becomes considerably low. On the other hand.

When starting multiple jobs at the same time, as you increase the multiplicity (number of jobs that can be started at once), the CPU usage rate (ratio of CPU usage time to total processing time) will eventually decrease to about 100. %become. This time is the optimum state with the highest processing efficiency.

If the degree of multiplicity is further increased, paging occurs frequently due to memory shortage, resulting in an overload state in which processing time becomes longer due to waiting for paging I10.

For this reason, there is not much problem when starting multiple jobs of a certain size, but when starting jobs of different sizes, the load is small when one job is started, but the load on other jobs is small. When a certain job was started, it would become overloaded and the processing time would take longer.

Furthermore, conventionally, as a means to condition the start of two individual jobs, a deadline time scheduling function that specifies the execution end time of a job, a dependent job control function that specifies the execution order between specific jobs, etc. have been used. Ta.

[Problem that the invention seeks to solve]

In conventional job startup control methods, when the system's job startup control mechanism (scheduler) starts multiple jobs in the pending queue in parallel, it only needs jobs within the number of job spaces that can be managed simultaneously. If there was a job for which resources such as devices, volumes, and datasets could be allocated, the job with the highest priority was automatically selected and started.

In addition, at this time, even if the deadline time schedule function or dependent job control function is used, the time required to start each job is undefined, so the CP
It was not possible to perform job start-up control to a sufficient degree taking U's processing efficiency into consideration.

[Means for solving problems]

The present invention is useful when starting multiple jobs simultaneously.

The purpose is to provide a means to enable fine-grained scheduling control that takes into account CPU processing efficiency, and therefore, it provides a means for specifying the starting order among a plurality of specified jobs, and a means for specifying the starting order of specified jobs. For this, there is a means to condition the completion of the start of the job currently being executed,
Regarding the activation of a specified job, means is provided to wait for a predetermined period of time from the activation of the immediately preceding job, and to specify a combination of these to enable optimal job scheduling control.

[Effect]

According to the present invention, it is possible to easily set a group of jobs to be executed together at any given time, or a job or a group of jobs that are prohibited from being executed together.
Optimum job starting control can be performed to prevent too much or too little load from being applied to the CPU.

〔Example〕

FIG. 1 is a block diagram of a system according to an embodiment of the present invention.

In the figure, 1 is job schedule information, 2 is a job queue, 3 is a scheduler, and 4 is an analysis unit.

5 represents a job starting unit, 6.7 represents an inishake, and 8.9 represents a processor.

Job schedule information 1 is information regarding job startup control based on the present invention, and is registered in the system as part of a so-called procedure. Further, the plurality of input jobs are registered on a magnetic disk device (DASD).

Scheduler 3 reads job schedule information 1,
The analysis unit 4 analyzes the control statement and determines the startup control timing for the specified job.

Specifically, the sequential startup job name that can be multiprocessed.

Specify that the subsequent job will wait for startup for a predetermined period of time, and specify the conditions for completing the startup of the immediately preceding job for the subsequent job.

Schedule ends, etc.

FIG. 2 shows an example of job schedule information 1. 5TEPI to 5TEP8 is multiple job JO
It represents a control sequence for starting BOI or JOBO6. To explain briefly.

5TEP 1 starts job JOBOI first.

At 5TEP2, job JOB, 02 is started following JOBOI, and jobs JOBOI, JOBO2 are multi-processed.

5TEP3, the wait designation "*" prohibits starting of subsequent jobs for 600 seconds.

After 600 seconds, 5TEP4, 5TEP5, 5TEP6
Then, jobs JOBO3, JOB04, JOBO respectively.
5 and perform multiple processing.

In 5TEP6, the previous job JOBO5 is
Prohibits the startup of subsequent jobs until the startup of the job is completed.

At 5TEP7, job JOBO6 is started upon completion of start-up of job JOBO5.

At 5TEP8, the schedule ends.

The time from the start of startup to the completion of startup for JOBOL alone is T (JOBO1), and JOBOI.

The simultaneous (multiple) execution time of JOBO2,... is T (J
OBOl, JOBO2, -''), the multiplicity of the activated job is.

T (JOBO1) + T (JOBO2) +...<T
(JOBOI, JOBO2,...) Set so that it does not occur.

If the waiting time "*" is expressed as Ti.

Set as Ti:5T (JOBOI, JOBO2). This means that even if you start JOBO3 at the same time,
BOL and JOBO2 have almost finished starting, and the CP
This is to prevent U from being overloaded.

The condition "CK" is used to further ensure this, and is applied, for example, when the startup completion time of the immediately preceding job is uncertain.

These job schedule information are set based on actual measurement data or trial and error.

〔Effect of the invention〕

According to the present invention, an environment close to optimal state B (a state in which the total processing time is equal to the processing time of the CPU) is set using job schedule information.

It is possible to automatically start a user job and shorten the start processing time. It is also possible to prioritize emergency operations.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of a system according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of job schedule information. FIG. 3 is a timing diagram of an example of multiple job activation processing. In FIG. 1, l represents job schedule information. 2 is a job queue, 3 is a scheduler, 4 is an analysis section, 5
is the job startup part, and 6.7 is the inishake. 8.9 represents a processor.

Claims (1)

[Claims] A data processing system equipped with a function of multiplexly executing a plurality of jobs, comprising: job schedule information means for instructing the starting order of the plurality of jobs; The job schedule information means includes information specifying the starting order among the plurality of jobs to be started, as well as specifying completion of starting of the most recently started job as the starting condition. and information specifying waiting a predetermined time from the start of the previous job as a start condition, as necessary, and the scheduler means analyzes each of the above specified information in the job schedule information to A multiple job startup control method characterized by controlling job startup timing.